Mirendil lands $100M+ Google Cloud deal to scale self-improving AI
Exclusive: $1B-valued startup secures TPU and GPU access for recursive AI research.
AI lab Mirendil has signed a multiyear partnership with Google Cloud worth upward of $100 million, TechCrunch exclusively learned. The deal gives the startup access to Google's TPUs, Nvidia GPUs, and managed training clusters to support its self-improving AI research. Mirendil co-founder and CEO Behnam Neyshabur confirmed the value, noting it's roughly half of what Mirendil raised in seed funding at a $1 billion valuation in late June. The startup's goal is to develop recursive self-improving AI systems that can iteratively enhance themselves, eventually handling the work of an entire frontier AI lab. Co-founders previously worked at Anthropic, which has been exploring similar concepts.
Training self-improving AI requires enormous compute, and Mirendil's approach focuses on matching workloads to the right hardware. Co-founder Harsh Mehta emphasized that Google's diverse chip options allow Mirendil to mix and match accelerators, lowering costs for both the startup and its customers. Google's SVP Amin Vahdat framed the partnership as an example of orchestrating systems rather than just chip performance. In return, Google gains a strategic partner building frontier recursive AI that could be spun into enterprise offerings later.
- Multiyear Google Cloud deal worth over $100M, half of Mirendil's seed raise
- Access to Google TPUs, Nvidia GPUs, and managed training clusters
- Startup builds recursive self-improving AI to automate research; founders are ex-Anthropic
Why It Matters
Secure compute deals like this accelerate recursive AI development, reshaping research and enterprise AI capabilities.